1
$\begingroup$

Disclaimer 1: There were other questions related to the broad nature of our help-center in regard to what is on-topic, none of them have a good or clearly answer, in my opinion.

Data Science in its nature is a broad subject with applications ranging from NLP and Computer Vision to Predicting Stock Prices and Future Revenue, so it is quite difficult to pin-point what is and what is not on topic as basically anything can benefit from Data Science techniques.

Images are 2D Digital Signals, and Image Processing is clearly a subtopic of Signal Processing SE but with relevant influence on Computer Vision which is one of the most successful Deep Learning Application.

As basically anything can be related to DS and as the help center is pretty broad about what is on-topic on DS SE how should people define where to ask?

See for example this question from Dawny33♦, a high reputation and clearly an important member of our community:

  • The question is: What does 'energy' in image processing mean?

It points to a Data Science related article Seam Carving for Content-Aware Image Resizing, but the question itself is purely about image processing, a subset of Digital Signal Processing.

As Cross Validated covers machine learning models them selves and Signal Processing covers things like digital image processing it would be on-topic to have pure machine learning or pure digital image processing question on Data Science Stack Exchange?

Or should we accept this kind of question once Data Scientist themselves are problem solvers and might have to tackle outside their field of expertise to solve Data Science problems without the need to contextualize the subject inside Data Science? (One can ask about DIP to understand it in order to later apply that to a DS system)

$\endgroup$
  • $\begingroup$ I'm confused about your second-to-last question. Putting a question on hold doesn't damage anyone's reputation. I wonder if there might be a misunderstanding? $\endgroup$ – D.W. Apr 8 at 15:46
  • $\begingroup$ Even before been put on hold people might vote-down a question by simply thinking it is off-topic while is a explict vote feature for off topic question would help to identify it. $\endgroup$ – Pedro Henrique Monforte Apr 8 at 15:49
  • $\begingroup$ That can't be prevented. I don't think there's any way to implement such a feature (since the computer can't read your mind and tell why you are downvoting), and I don't think there's any chance of it being implemented. So, I would personally suggest removing that, to avoid derailing the discussion. We already have an "explicit vote for off-topic"; it's voting to close the question. $\endgroup$ – D.W. Apr 8 at 15:50
  • $\begingroup$ The propose is to have a option of downvote explicit for off-topic só people can express their reason $\endgroup$ – Pedro Henrique Monforte Apr 8 at 15:52
  • $\begingroup$ I know you say you want to discuss all of these together, but I would recommend separating out that discussion: one discussion of whether it is on-topic vs off-topic, a separate discussion about what to do about it (e.g., buttons downvoting because it is off-topic). $\endgroup$ – D.W. Apr 8 at 15:54
  • $\begingroup$ You are right, I will do so. $\endgroup$ – Pedro Henrique Monforte Apr 8 at 16:10
0
$\begingroup$

My immediate instinct is to suggest that image processing be off-topic. Why? It seems pretty far afield from data science to me. Also, it seems adequately covered on other sites (both Stack Overflow and Signal Processing).

In contrast, applications of machine learning to computer vision should be on-topic.


A different way to consider it would be: Does the question need to be answered from a data science perspective? Would data scientists have unique expertise to answer it, that isn't available in other communities? Or, is the question of unique relevance and interest to data scientists? If the answers are yes, then it seems suitable here. However, a random question about image processing typically isn't something that needs to be answered from a data science perspective and so wouldn't fall into that category.

$\endgroup$
  • $\begingroup$ The second part of your answer, I think that defines on-topic better than the examples on the help center. We should formalize the help center in some way, examples are too broad and open to interpretation. $\endgroup$ – Pedro Henrique Monforte Apr 9 at 22:05
2
$\begingroup$

Image processing is a topic which is related to Computer Vision. In fact, most of the times, they are used interchangeably, not saying that that's right.

If you look at the question, the OP asks it from the perspective of a paper related to Computer Vision. In this case, the OP clearly doesn't know what 'energy' means in the context of the paper. So, they'd never be able to figure out if the question is related to image processing or CV.

So, to answer your question: On-topic: Image Processing?

Yes. They are on-topic as long as the perspective/context is related to the topics which are on-topic in this site (in most cases Computer Vision)

$\endgroup$
  • $\begingroup$ Still, Computer Vision is a topic related but not entirely covered in Data Science. $\endgroup$ – Pedro Henrique Monforte Apr 8 at 17:21
  • $\begingroup$ Actually this is exactly my point when I say "Or should we accept this kind of question once Data Scientist themselves are problem solvers and might have to tackle outside their field of expertise to solve Data Science problems without the need to contextualize the subject inside Data Science? (One can ask about DIP to understand it in order to later apply that to a DS system)". $\endgroup$ – Pedro Henrique Monforte Apr 8 at 17:23
  • $\begingroup$ Your point of Computer Vision is a topic related but not entirely covered in Data Science is debatable. And I respectfully disagree to that statement $\endgroup$ – Dawny33 Apr 8 at 17:24
  • $\begingroup$ If you are not someone from Digital Signal Processing you are not supposed to know that "energy" in image processing is completly image processing and has little to do with DS. Also Energy is used into superpixels, which is an old image processing technique reliant on clustering. $\endgroup$ – Pedro Henrique Monforte Apr 8 at 17:25
  • $\begingroup$ Image Registering for example. This is clearly a Computer Vision technique and have solutions using Machine Learning but there are solutions based on methods that are not learned but rather manually defined. $\endgroup$ – Pedro Henrique Monforte Apr 8 at 17:26
  • $\begingroup$ I never disagreed with you on that. As I mentioned, it is about the context. Here in the qn quoted, the context was CV. + just because something is on-topic on a site doesnt make that off-topic on another site :) $\endgroup$ – Dawny33 Apr 8 at 17:26
  • $\begingroup$ This is true, this particular question lays in a gray area. It would be best answered by Signal Processing SE members, is it okay to ask exactly the same question in two stack exchange communities? $\endgroup$ – Pedro Henrique Monforte Apr 8 at 17:28
  • 2
    $\begingroup$ is it okay to ask exactly the same question in two stack exchange communities?. No. It creates duplicates and often frowned upon. Ask in one of the either. $\endgroup$ – Dawny33 Apr 8 at 17:30
  • $\begingroup$ I see three topics: use of machine learning for computer vision; computer vision more generally; image processing. I think it is possible to distinguish between these three; just because the first is on-topic doesn't mean the second or third have to be on-topic. I think the use of machine learning to solve computer vision problems) should be on-topic here. (continued) $\endgroup$ – D.W. Apr 8 at 20:41
  • 1
    $\begingroup$ I'm not convinced that all of computer vision should be on-topic here; for computer vision algorithms that don't use any machine learning and don't involve learning from large data sets, I'm not seeing how that is part of data science. Image processing is a step even further away. Your argument is "image processing is too hard to distinguish from computer vision, and computer vision should be on-topic, so image processing should be too". I'm not sure I agree with either of those premises -- especially the second one, that all computer vision questions should be considered on-topic here. $\endgroup$ – D.W. Apr 8 at 20:41
2
$\begingroup$

This is one of those gray areas. If there is a dedicated SE for a topic, it's probably the best place for that question in almost all cases. However, I can imagine a question being on-topic in both if it concerns, say, training an ML model to recognize objects in an image. I will generally not close as off-topic if it's arguably also relevant here. If it's much more on-topic elsewhere i'd migrate it.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .